Many robot projects fail before the robot arrives. They start with an impressive demo, a purchase and a search for something useful to do with the machine. Months later it stands in a corner. The projects that succeed start the other way round: with one clearly chosen piece of work. Here is how to find it and how to run it.

Step 1: List the candidates

Walk the floor with the people who do the work and write down every task that is repetitive, physically tiring or hard to staff. Keep every idea for now. Typical candidates are moving trolleys between areas, feeding parts to a line, loading machines, moving empty containers back and carrying finished goods to shipping.

Step 2: Score each task

Rate each candidate from one to five on these points:

  • Volume: how many times per shift does it happen?
  • Repetition: is it done the same way each time?
  • Variation: how many different parts, routes or exceptions are involved? Fewer is better.
  • Strain: is it heavy, awkward or unpopular with staff?
  • Data: is the task already recorded in a system such as SAP, so a robot can receive it automatically?
  • Space: is there room for a robot to move safely?

The best first task scores high on volume, repetition, strain and data, and low on variation. It should be important enough to matter, while a bad day on it still leaves the rest of the plant running.

Step 3: Measure today

Before anything changes, measure the current process for at least a few weeks: how long each task takes, how many people hours it uses per shift, how far people walk, how often things go wrong and what the knock-on effects are. This baseline is what lets you prove the robot made a difference.

Step 4: Define success in advance

Agree on a short list of numbers that will decide whether the pilot worked, for example tasks completed per shift, share of tasks finished without human help, and time people get back for other work. Write them down and agree them with management, operations and the team before the pilot starts.

Step 5: Bring your people in early

The operators who do the job today are your most valuable source of knowledge. Involve them in choosing the task, designing the route and testing the robot. Explain what will change for them and what will stay the same.

In Germany and many other European countries, involve the works council early as well. Introducing technology that can record how work is done is usually subject to co-determination, and a works council that has been part of the project from the start becomes a strong ally.

Step 6: Plan safety and integration

Two pieces of work run in parallel with the preparation:

  • Safety: a risk assessment for the new setup, covering routes, speeds, crossings with people and forklifts, and what happens in an emergency.
  • Integration: how the robot receives its tasks, how it reports them as done and what happens when it cannot finish one. A robot that needs someone to type in every job saves very little time.

Step 7: Run the pilot in phases

  • Phase one: the robot works in a limited area, at limited times, with a person close by.
  • Phase two: the area and hours grow, and the person steps back to monitoring.
  • Phase three: the robot runs the task as part of normal operations, and the team handles exceptions.

Review the numbers from step 4 at the end of each phase and only move on when they hold.

Step 8: Decide and scale

At the end of the pilot, compare the results with the baseline and the goals you set. If they hold, scale the same task to more shifts or areas before adding new tasks. Each new workflow is faster than the last, because the connection to your systems, the safety concept and the experience of your team are already in place.

Common mistakes to avoid

  • Choosing the most impressive task over the most useful one.
  • Skipping the baseline, so success cannot be proven.
  • Leaving operators and the works council out until the robot arrives.
  • Treating the robot as a stand-alone gadget, separate from the process.
  • Trying to automate five tasks at once.

Start with one workflow, do it properly and let the results make the case for the next one.

Talk to us about your first workflow